Observational studies - basics, agreement with randomized trials, flawed vaccine studies, UCSF talk
Vložit
- čas přidán 2. 11. 2023
- Vinay Prasad, MD MPH; Physician & Professor
Hematologist/ Oncologist
Professor of Epidemiology, Biostatistics and Medicine
Author of 450+ Peer Reviewed papers, 2 Books, 2 Podcasts, 100+ op-eds.
If you want to contact me, do it here: www.vinayakkprasad.com/contact
Google Scholar: scholar.google.com/citations?...
Substack: vinayprasadmdmph.substack.com/
Podcast: podcasts.apple.com/us/podcast...
Personal Website: www.vinayakkprasad.com
Laboratory Website: www.vkprasadlab.com
Podcast Website: www.plenarysessionpodcast.com
Academic Publications: www.vinayakkprasad.com/papers
Follow me on:
Twitter @vprasadmdmph
I am reading your book Malignant. What a great book! What an eye opener.
Incentives and deception must also be considered for validity of observation or studies or peer review. Fraud mitigation needs to be part of all these processes. We're seen the very real institutional capture by industry and ideology the last few years and something like separation of powers and funding of unpopular ideas by appropriately designed studies that aren't designed to fail (such as by incorrect dosage or improper timing). Critics must be involved in the study design or their concerns won't be addressed and nothing will have been solved because they won't move on and the debate at an impass.
Or we just reject bad studies as evidence.
Nailed it Brohahn
@@DylanYoungif you have total visibility of all process and data, then rejecting bad studies is simple. The only way for that to happen is for third parties with totally different incentives to run ALL studies.
This is exactly the type of content that I’d like to see more of.
I have been watching and listening to you for about 3 years now. You need to put together a set of videos aimed at the high school literacy level. Your information is great but lost in translation for those with limited understanding of immunology, disease and basic research methodology. Not a criticism just that your level o knowledge is way above most people’s everyday understanding. I am high school biology teacher with 10 years experience in vaccine research before I decided to become a teacher. Most high school teachers don’t have a clue on how real research is carried out. This is a much needed resource.
Great point, Jason. How about this for another shocking fact- MOST medical students and physicians "don't have a clue on how real research is carried out."
The number of "researchers" I've met with no understanding of the statistics they're using is staggering.
@@JMK-vo8pvthis. Todays “doctors” just regurgitate what some other ninny told them. Critical thinking has left the medical community!!
Bless yer heart, honey, but those doctors ain't always the sharpest tools in the shed. They spend all their time memorizin' big fancy words and lookin' at pictures of insides, but they don't know much 'bout real-life medicine. Now, I'm not sayin' they're all bad. There's a few good ones out there, but most of 'em are just followin' the herd. They see what the other doctors are doin' and they figure it must be the right thing to do, even if it don't make no sense. And don't even get me started on those guidelines they always talkin' about. Those things are written by a bunch of folks who got their pockets lined by the big pharma companies. They don't care about what's best for you, they just care about sellin' their drugs. So next time you go to the doctor, take a deep breath and hold on tight. You're in for a wild ride.
Them doctors sometimes ain't as versed in statistics as a raccoon in a cornfield. It's like tryin' to teach a bullfrog to two-step - they might be good at doctorin', but when it comes to numbers, they're like a chicken tryin' to play the banjo. Them doctors, darlin', they're followin' guidelines like a hound dog on a scent, but them guidelines? Written by folks with more conflicts of interest than a henhouse with a fox in it. It's a regular old hootenanny of pharmaceutical dollars and doctorin' decisions. It's like them doctors are playin' with Legos - stackin' up them medications without always thinkin' 'bout the big picture. Bless their hearts, they mean well, but sometimes it's like askin' a pig to recite Shakespeare - it just ain't in their wheelhouse. So, next time you're sittin' in that doctor's office, just remember, they might be good at patchin' you up, but when it comes to statistics and evidence-based medicine, well, it's like teachin' a bull to ballet - it takes a bit of doin'.
Thank you.
I remain convinced that the simple change from 95% CLs to 99.7% CLs would make an massive improvement.
We are hiding our ignorance of actual causal mechanisms between 2 and 3 sigma.
Much respect for what you're doing.be careful, these ppl are whack jobs. They love taking ppl out, no problem. Please (I know from experience) be careful, you first. Remember the many millions already sick, injured and unalive. We don't matter to them. But we do to each other. Thank you for being brave, stay safe.
Thank you Dr Prasad.
Do you know why Professor Thomas Seyfried's treatment for cancer hasn't been adopted?
It seems logical and very effective.
Is it cheap and unpatentable?
@@mballer yep, I hope that's not the reason, it's very sad if it is.
Great talk!
If Nothing Changes, Nothing Changes.
Is there something profound in your post? I missed your point.
Why isn't Bayesian inference used more often? It seems a much more applicable and effective tool in our world of imperfect and always updating knowledge than after-the-fact collection and deduction with frequentist statistical techniques and calculations. Don't these results give a false impression of precise and definite knowledge?
In my experience, Bayesian inference is simply not taught within “typical” public health curricula. Usually it’s really good schools that could cover it and you still kinda have to seek it out on your own. After navigating around this space for a while I’ve noticed that most public health people are not really well versed in calculus so those crazy integrals that show up in the denominator get unappetizing quite quickly.
1) Ignorance or malice.
2) Yes.
Aaaaah Vinay!!! Don’t skip the parachute analogy. Thank you for your analysis and clear vision of how science should be approached.
Excellent lecture as usual. Inspiring job by Dr Prasad. I just wanna ask if someone can explain me a little bit more about de time zero effect in observational studies. Thank you.
well presented.thanks
Excellent, helpful, relevant. Thanks
If the full boimechanical mechanism of action is understood and the "intervention " fits logically into that pathway in a known way, no experiment is needed except for side effects and study of other unknowns. This is called modelling. A parachute and cigarettes (by known carcinogens) both fit fully into our current knowledge base.
How do you adjust this for genetic therapies in ultra rare diseases
41:00 err checking. (driving test) _JC
I think the largest issue is that you have a patient with a very bleak outlook and you know (just like the guy jumping out of the airplane) if you do NOTHING it will end badly. "Do nothing" doesn't cure cancer.
So people try whatever might work.
Its good we have best standard practice therapies, so we can compare with those - but did all of them come about from random trials? Or mostly from people looking at what seems to work.
I swear I listened to every word but totally admiring the long hair look lol.
Per usual, this was an outstanding lecture. However, the question and answer phase of the video was very hard to follow because we cannot hear the questions coming from the audience AND the captioning didn't sych up well with the questions being asked.
Can you do a podcast on any improvements in Hematology? I have to get Ferritin infusions as I just don’t make or store my own. I’ve started having reactions to Venofer. So what are other options?
Cochrane Library’s strong acknowledgement of the value of observational control studies: Cochrane Systematic Review from data of 1583 meta-analyses that covered 228 different medical conditions.: “Healthcare outcomes assessed with observational study designs compared with those assessed in randomized trials” The Cochrane Library determined: “Our results provide little evidence for significant effect estimate differences between observational studies and RCTs, regardless of specific observational study design, heterogeneity, inclusion of pharmacological studies, or use of propensity score adjustment”.
As far as putting things in streams, you still need to look at cost and benefit. If the benefit is large compared to the cost, say, someone proposed at least 10x, them you may want to allow it.
Towards the beginning of the video, something very important was said: RCTs are not conducted to test for adverse effects or harm. Although understandable, it provides promoters of interventions with a solid defence against criticism : causal evidence of harmfulness is often weak.
55:40 Randomized ~= auto cure all. _JC
Very logical lecture with great slides and excellent presentation skills.
Thanks Dr Prasad for taking the time to make the scientific vocabulary more easier for the lay person to understand using all the analogies thrown in. Your conclusions stemming from statistical drawbacks/ discrepancies in current medical interventions is very sound logic and ethically based.
Please continue what you do best but on a different platform due to the new legislation taking effect.
Take Care !
PS Time for a haircut, young man.
Whole Hardily disagree on everything you said in that comment
Regarding hair, i mean
15:30 provocative publications. _JC
I remain confused about why we'd actually study things when we could just be ideological about them, which would be so much cheaper
What is being "ideological"?
@@mballeroperating on a "belief" system.
Lol
Just as pharmaceutical companies practice!
@@lisaBinfinity yes.
But then, why was the observational study of breast surgery the wrong one or with unreliable results?
I'm guessing he considers it to be much more confounded than the randomized trial. There's probably some bias similar to the chemotherapy example he shared, where the people receiving the intervention are healthier or something like that. Just an idea.
Randomized studies don't look at all the variables
There is no randomized vs oberservational, they are both flawed.
If the effect is very large, either study type would be fine.
If the effect is small neither study type is reliable.
Reverse engineering studies are the best.
Randomised control trials utilize the law of large random numbers to neglect variables. It's true that it's not a guarantee that confounding factors don't exist. Reverse engineering has flaws too (often we think we know how something works, but we're wrong).
@@DylanYoung
I guess I have my own definition of reverse engineering that isn't yours.
My reverse engineering records every minute detail of ones life, not just a drug and a few variables.
Every cheeseburger fries and coke is recorded.
Every step of every day is recorded.
All blood work is recorded.
If a drug looks like it worked for someone all the details of that person's life is compared to others who also were successful with a drug.
Commonalities are calculated.
It's not a money maker.